|
|
Created:
6 years, 7 months ago by Erik Corry Modified:
6 years, 7 months ago CC:
blink-reviews, haraken, kouhei+heap_chromium.org, Mads Ager (chromium) Base URL:
svn://svn.chromium.org/blink/trunk Visibility:
Public. |
DescriptionSimplify and speed up address-to-page cache for conservative stack scanning.
This change makes the per-heap address-to-page cache positive only and handles
large pages more naturally. There is now a separate whole-process negative
cache that keeps track of pages that are not in any of the heaps. This speeds up my microbenchmark from ~87 to ~65.
R=ager@chromium.org, haraken@chromium.org, wibling@chromium.org, zerny@chromium.org
BUG=
Committed: https://src.chromium.org/viewvc/blink?view=rev&revision=173749
Patch Set 1 #
Total comments: 34
Patch Set 2 : Revert GC heuristic change and CL feedback #Patch Set 3 : There were some missing cache flushes. Strengthened asserts #
Total comments: 7
Patch Set 4 : Merge up #
Messages
Total messages: 17 (0 generated)
LGTM with comments. This is a nice change! https://codereview.chromium.org/271703002/diff/1/Source/platform/heap/Heap.cpp File Source/platform/heap/Heap.cpp (right): https://codereview.chromium.org/271703002/diff/1/Source/platform/heap/Heap.cp... Source/platform/heap/Heap.cpp:763: heapContainsCache()->flush(); For consistency, let's define flushHeapContainsCache(). https://codereview.chromium.org/271703002/diff/1/Source/platform/heap/Heap.cp... Source/platform/heap/Heap.cpp:837: heapContainsCache()->flush(); flushHeapContainsCache() https://codereview.chromium.org/271703002/diff/1/Source/platform/heap/Heap.cp... Source/platform/heap/Heap.cpp:1529: if (reinterpret_cast<uintptr_t>(address) < blinkPageSize) What is this change for? https://codereview.chromium.org/271703002/diff/1/Source/platform/heap/Heap.h File Source/platform/heap/Heap.h (right): https://codereview.chromium.org/271703002/diff/1/Source/platform/heap/Heap.h#... Source/platform/heap/Heap.h:192: virtual bool isLargeObject() { return true; } Add OVERRIDE. https://codereview.chromium.org/271703002/diff/1/Source/platform/heap/Heap.h#... Source/platform/heap/Heap.h:241: return ((size() - 1) & blinkPageBaseMask) + blinkPageSize; Can we use roundToBlinkPageEnd? If you use roundToBlinkPageEnd in the caller site, you can remove sizeRoundedUpToPage. https://codereview.chromium.org/271703002/diff/1/Source/platform/heap/Heap.h#... Source/platform/heap/Heap.h:519: AddressEntry(Address address) : m_address(address) { } Add explicit. https://codereview.chromium.org/271703002/diff/1/Source/platform/heap/Heap.h#... Source/platform/heap/Heap.h:553: NegativeEntry(Address address, bool) : AddressEntry(address) { } Do we need the bool parameter? NegativeEntry should just contain a list of negative entries, and boolean parameters won't need to be associated. https://codereview.chromium.org/271703002/diff/1/Source/platform/heap/Heap.h#... Source/platform/heap/Heap.h:901: static void flushNotInHeapCache(); flushHeapDoesNotContainCache It's confusing to have two names for one thing. https://codereview.chromium.org/271703002/diff/1/Source/platform/heap/Heap.h#... Source/platform/heap/Heap.h:908: static HeapDoesNotContainCache* s_notInHeapCache; s_notInHeapCache => s_heapDoesNotContainCache https://codereview.chromium.org/271703002/diff/1/Source/platform/heap/ThreadS... File Source/platform/heap/ThreadState.cpp (right): https://codereview.chromium.org/271703002/diff/1/Source/platform/heap/ThreadS... Source/platform/heap/ThreadState.cpp:525: if (newSize < (1 << 20)) What is this condition for? https://codereview.chromium.org/271703002/diff/1/Source/platform/heap/ThreadS... Source/platform/heap/ThreadState.cpp:664: ASSERT(!page || page->isLargeObject() || page == m_heaps[i]->heapPageFromAddress(blinkPageAddr)); heapPageFromAddress(roundToBlinkPageStart(address)) https://codereview.chromium.org/271703002/diff/1/Source/platform/heap/ThreadS... Source/platform/heap/ThreadState.cpp:665: ASSERT(!page || page->isLargeObject() || page == m_heaps[i]->heapPageFromAddress(blinkPageAddr + blinkPageSize - 1)); heapPageFromAddress(roundToBlinkPageEnd(address)) https://codereview.chromium.org/271703002/diff/1/Source/platform/heap/ThreadS... Source/platform/heap/ThreadState.cpp:666: if (page) { You can add the two ASSERTs into this if branch. Then you won't need to check !page.
Nothing to add to Haraken's comments, lgtm
LGTM https://codereview.chromium.org/271703002/diff/1/Source/platform/heap/Heap.cpp File Source/platform/heap/Heap.cpp (right): https://codereview.chromium.org/271703002/diff/1/Source/platform/heap/Heap.cp... Source/platform/heap/Heap.cpp:1529: if (reinterpret_cast<uintptr_t>(address) < blinkPageSize) This is an optimization to quickly filter out small integers (which occur pretty frequently in stack frames). Could you add a comment stating that? https://codereview.chromium.org/271703002/diff/1/Source/platform/heap/Heap.h File Source/platform/heap/Heap.h (right): https://codereview.chromium.org/271703002/diff/1/Source/platform/heap/Heap.h#... Source/platform/heap/Heap.h:900: static void addressIsNotInHeap(Address); I would remove these methods and just inline the one liners. I find that easier to read (being explicit about the cache) and it doesn't clutter the header file. Are they added because ThreadState needs to use this? ThreadState is a friend and has access to the cache directly. You could add a private getter if you want. https://codereview.chromium.org/271703002/diff/1/Source/platform/heap/ThreadS... File Source/platform/heap/ThreadState.cpp (right): https://codereview.chromium.org/271703002/diff/1/Source/platform/heap/ThreadS... Source/platform/heap/ThreadState.cpp:525: if (newSize < (1 << 20)) Could we do tuning of the GC heuristics separately from this change? Let's introduce the negative cache first. Then we can tune these as a follow-up (and update the comments when we do.)
lgtm https://codereview.chromium.org/271703002/diff/1/Source/platform/heap/ThreadS... File Source/platform/heap/ThreadState.cpp (right): https://codereview.chromium.org/271703002/diff/1/Source/platform/heap/ThreadS... Source/platform/heap/ThreadState.cpp:475: BaseHeapPage* page = heapPageFromAddress(address); NIT: Perhaps add a comment that this will check both "normal" and large objects since the variable name doesn't indicate the latter.
https://codereview.chromium.org/271703002/diff/1/Source/platform/heap/Heap.cpp File Source/platform/heap/Heap.cpp (right): https://codereview.chromium.org/271703002/diff/1/Source/platform/heap/Heap.cp... Source/platform/heap/Heap.cpp:763: heapContainsCache()->flush(); On 2014/05/08 05:44:58, haraken wrote: > > For consistency, let's define flushHeapContainsCache(). Done, but it doesn't make it very consistent with flushNotInHeapCache. flushNotInHeapCache is a static method on Heap, whereas flushHeapContainsCache is an instance method on ThreadHeap. https://codereview.chromium.org/271703002/diff/1/Source/platform/heap/Heap.cp... Source/platform/heap/Heap.cpp:837: heapContainsCache()->flush(); On 2014/05/08 05:44:58, haraken wrote: > > flushHeapContainsCache() Done. https://codereview.chromium.org/271703002/diff/1/Source/platform/heap/Heap.cp... Source/platform/heap/Heap.cpp:1529: if (reinterpret_cast<uintptr_t>(address) < blinkPageSize) On 2014/05/08 05:44:58, haraken wrote: > > What is this change for? You can't allocate at address 0, so no allocation can take place in the first page (pages are aligned). I moved the negative cache lookup here instead of having it in the ThreadState. That makes this test unnecessary. https://codereview.chromium.org/271703002/diff/1/Source/platform/heap/Heap.cp... Source/platform/heap/Heap.cpp:1529: if (reinterpret_cast<uintptr_t>(address) < blinkPageSize) On 2014/05/08 06:52:37, Mads Ager (chromium) wrote: > This is an optimization to quickly filter out small integers (which occur pretty > frequently in stack frames). Could you add a comment stating that? Removed instead. https://codereview.chromium.org/271703002/diff/1/Source/platform/heap/Heap.h File Source/platform/heap/Heap.h (right): https://codereview.chromium.org/271703002/diff/1/Source/platform/heap/Heap.h#... Source/platform/heap/Heap.h:192: virtual bool isLargeObject() { return true; } On 2014/05/08 05:44:58, haraken wrote: > > Add OVERRIDE. Done. https://codereview.chromium.org/271703002/diff/1/Source/platform/heap/Heap.h#... Source/platform/heap/Heap.h:241: return ((size() - 1) & blinkPageBaseMask) + blinkPageSize; On 2014/05/08 05:44:58, haraken wrote: > > Can we use roundToBlinkPageEnd? If you use roundToBlinkPageEnd in the caller > site, you can remove sizeRoundedUpToPage. Removed https://codereview.chromium.org/271703002/diff/1/Source/platform/heap/Heap.h#... Source/platform/heap/Heap.h:519: AddressEntry(Address address) : m_address(address) { } On 2014/05/08 05:44:58, haraken wrote: > > Add explicit. Done. https://codereview.chromium.org/271703002/diff/1/Source/platform/heap/Heap.h#... Source/platform/heap/Heap.h:553: NegativeEntry(Address address, bool) : AddressEntry(address) { } On 2014/05/08 05:44:58, haraken wrote: > > Do we need the bool parameter? NegativeEntry should just contain a list of > negative entries, and boolean parameters won't need to be associated. The addEntry method of HeapExtentCache expects to use two arguments. https://codereview.chromium.org/271703002/diff/1/Source/platform/heap/Heap.h#... Source/platform/heap/Heap.h:900: static void addressIsNotInHeap(Address); On 2014/05/08 06:52:37, Mads Ager (chromium) wrote: > I would remove these methods and just inline the one liners. I find that easier > to read (being explicit about the cache) and it doesn't clutter the header file. > > > Are they added because ThreadState needs to use this? ThreadState is a friend > and has access to the cache directly. You could add a private getter if you > want. Done. https://codereview.chromium.org/271703002/diff/1/Source/platform/heap/Heap.h#... Source/platform/heap/Heap.h:901: static void flushNotInHeapCache(); On 2014/05/08 05:44:58, haraken wrote: > > flushHeapDoesNotContainCache > > It's confusing to have two names for one thing. Done. https://codereview.chromium.org/271703002/diff/1/Source/platform/heap/Heap.h#... Source/platform/heap/Heap.h:908: static HeapDoesNotContainCache* s_notInHeapCache; On 2014/05/08 05:44:58, haraken wrote: > > s_notInHeapCache => s_heapDoesNotContainCache Done. https://codereview.chromium.org/271703002/diff/1/Source/platform/heap/ThreadS... File Source/platform/heap/ThreadState.cpp (right): https://codereview.chromium.org/271703002/diff/1/Source/platform/heap/ThreadS... Source/platform/heap/ThreadState.cpp:475: BaseHeapPage* page = heapPageFromAddress(address); On 2014/05/08 07:43:41, wibling-chromium wrote: > NIT: Perhaps add a comment that this will check both "normal" and large objects > since the variable name doesn't indicate the latter. Done. https://codereview.chromium.org/271703002/diff/1/Source/platform/heap/ThreadS... Source/platform/heap/ThreadState.cpp:525: if (newSize < (1 << 20)) On 2014/05/08 06:52:37, Mads Ager (chromium) wrote: > Could we do tuning of the GC heuristics separately from this change? Let's > introduce the negative cache first. Then we can tune these as a follow-up (and > update the comments when we do.) Done. https://codereview.chromium.org/271703002/diff/1/Source/platform/heap/ThreadS... Source/platform/heap/ThreadState.cpp:525: if (newSize < (1 << 20)) On 2014/05/08 05:44:58, haraken wrote: > > What is this condition for? Heuristic. Removed for now. https://codereview.chromium.org/271703002/diff/1/Source/platform/heap/ThreadS... Source/platform/heap/ThreadState.cpp:664: ASSERT(!page || page->isLargeObject() || page == m_heaps[i]->heapPageFromAddress(blinkPageAddr)); On 2014/05/08 05:44:58, haraken wrote: > > heapPageFromAddress(roundToBlinkPageStart(address)) Done. https://codereview.chromium.org/271703002/diff/1/Source/platform/heap/ThreadS... Source/platform/heap/ThreadState.cpp:665: ASSERT(!page || page->isLargeObject() || page == m_heaps[i]->heapPageFromAddress(blinkPageAddr + blinkPageSize - 1)); On 2014/05/08 05:44:58, haraken wrote: > > heapPageFromAddress(roundToBlinkPageEnd(address)) Done. https://codereview.chromium.org/271703002/diff/1/Source/platform/heap/ThreadS... Source/platform/heap/ThreadState.cpp:666: if (page) { On 2014/05/08 05:44:58, haraken wrote: > > You can add the two ASSERTs into this if branch. Then you won't need to check > !page. Done.
https://codereview.chromium.org/271703002/diff/40001/Source/platform/heap/Hea... File Source/platform/heap/Heap.cpp (right): https://codereview.chromium.org/271703002/diff/40001/Source/platform/heap/Hea... Source/platform/heap/Heap.cpp:1529: #ifdef NDEBUG Why not do this optimization in release builds? Is it a left-over from debugging? https://codereview.chromium.org/271703002/diff/40001/Source/platform/heap/Hea... Source/platform/heap/Heap.cpp:1546: #else Ditto? https://codereview.chromium.org/271703002/diff/40001/Source/platform/heap/Thr... File Source/platform/heap/ThreadState.cpp (right): https://codereview.chromium.org/271703002/diff/40001/Source/platform/heap/Thr... Source/platform/heap/ThreadState.cpp:479: // other other heaps or put this address in the other other -> other https://codereview.chromium.org/271703002/diff/40001/Source/platform/heap/Thr... Source/platform/heap/ThreadState.cpp:656: #ifdef NDEBUG Should this be removed as well?
https://codereview.chromium.org/271703002/diff/40001/Source/platform/heap/Hea... File Source/platform/heap/Heap.cpp (right): https://codereview.chromium.org/271703002/diff/40001/Source/platform/heap/Hea... Source/platform/heap/Heap.cpp:1529: #ifdef NDEBUG On 2014/05/09 10:25:20, wibling-chromium wrote: > Why not do this optimization in release builds? Is it a left-over from > debugging? It's confusing because the macro has negative sense. So ifdef NDEBUG means if-we-are-not-debugging. The reason we don't do the optimization when we are debugging is that we want to assert the cache is working in debug mode. https://codereview.chromium.org/271703002/diff/40001/Source/platform/heap/Hea... Source/platform/heap/Heap.cpp:1546: #else On 2014/05/09 10:25:20, wibling-chromium wrote: > Ditto? Ditto
lgtm https://codereview.chromium.org/271703002/diff/40001/Source/platform/heap/Hea... File Source/platform/heap/Heap.cpp (right): https://codereview.chromium.org/271703002/diff/40001/Source/platform/heap/Hea... Source/platform/heap/Heap.cpp:1529: #ifdef NDEBUG On 2014/05/09 10:31:54, Erik Corry wrote: > On 2014/05/09 10:25:20, wibling-chromium wrote: > > Why not do this optimization in release builds? Is it a left-over from > > debugging? > > It's confusing because the macro has negative sense. So ifdef NDEBUG means > if-we-are-not-debugging. > > The reason we don't do the optimization when we are debugging is that we want to > assert the cache is working in debug mode. Doh, okay then:)
The CQ bit was checked by erik.corry@gmail.com
CQ is trying da patch. Follow status at https://chromium-status.appspot.com/cq/erik.corry@gmail.com/271703002/40001
The CQ bit was unchecked by commit-bot@chromium.org
Failed to apply patch for Source/platform/heap/Heap.h: While running patch -p1 --forward --force --no-backup-if-mismatch; patching file Source/platform/heap/Heap.h Hunk #1 succeeded at 90 (offset 2 lines). Hunk #2 succeeded at 141 (offset 2 lines). Hunk #3 succeeded at 159 (offset 2 lines). Hunk #4 succeeded at 191 (offset 2 lines). Hunk #5 succeeded at 214 (offset 2 lines). Hunk #6 succeeded at 255 (offset 2 lines). Hunk #7 succeeded at 457 (offset 2 lines). Hunk #8 succeeded at 486 (offset 2 lines). Hunk #9 FAILED at 508. Hunk #10 succeeded at 701 (offset 52 lines). Hunk #11 succeeded at 747 (offset 52 lines). Hunk #12 succeeded at 763 (offset 52 lines). Hunk #13 succeeded at 910 (offset 52 lines). 1 out of 13 hunks FAILED -- saving rejects to file Source/platform/heap/Heap.h.rej Patch: Source/platform/heap/Heap.h Index: Source/platform/heap/Heap.h diff --git a/Source/platform/heap/Heap.h b/Source/platform/heap/Heap.h index e464e89dcfb4c2596f1cc1f532ebc3f2c16f394c..3ed53923db6c589f346a340efd17fc424803eea7 100644 --- a/Source/platform/heap/Heap.h +++ b/Source/platform/heap/Heap.h @@ -88,6 +88,11 @@ inline Address roundToBlinkPageStart(Address address) return reinterpret_cast<Address>(reinterpret_cast<uintptr_t>(address) & blinkPageBaseMask); } +inline Address roundToBlinkPageEnd(Address address) +{ + return reinterpret_cast<Address>(reinterpret_cast<uintptr_t>(address - 1) & blinkPageBaseMask) + blinkPageSize; +} + // Compute the amount of padding we have to add to a header to make // the size of the header plus the padding a multiple of 8 bytes. template<typename Header> @@ -134,17 +139,15 @@ public: ASSERT(isPageHeaderAddress(reinterpret_cast<Address>(this))); } - // Check if the given address could point to an object in this + // Check if the given address points to an object in this // heap page. If so, find the start of that object and mark it - // using the given Visitor. - // - // Returns true if the object was found and marked, returns false - // otherwise. + // using the given Visitor. Otherwise do nothing. The pointer must + // be within the same aligned blinkPageSize as the this-pointer. // // This is used during conservative stack scanning to // conservatively mark all objects that could be referenced from // the stack. - virtual bool checkAndMarkPointer(Visitor*, Address) = 0; + virtual void checkAndMarkPointer(Visitor*, Address) = 0; #if ENABLE(GC_TRACING) virtual const GCInfo* findGCInfo(Address) = 0; @@ -154,6 +157,7 @@ public: PageMemory* storage() const { return m_storage; } ThreadState* threadState() const { return m_threadState; } const GCInfo* gcInfo() { return m_gcInfo; } + virtual bool isLargeObject() { return false; } private: // Accessor to silence unused warnings. @@ -185,11 +189,14 @@ public: COMPILE_ASSERT(!(sizeof(LargeHeapObject<Header>) & allocationMask), large_heap_object_header_misaligned); } - virtual bool checkAndMarkPointer(Visitor*, Address); + virtual void checkAndMarkPointer(Visitor*, Address) OVERRIDE; + virtual bool isLargeObject() OVERRIDE { return true; } #if ENABLE(GC_TRACING) - virtual const GCInfo* findGCInfo(Address) + virtual const GCInfo* findGCInfo(Address address) { + if (!objectContains(address)) + return 0; return gcInfo(); } #endif @@ -205,9 +212,19 @@ public: *previousNext = m_next; } + // The LargeHeapObject pseudo-page contains one actual object. Determine + // whether the pointer is within that object. + bool objectContains(Address object) + { + return (payload() <= object) && (object < address() + size()); + } + + // Returns true for any address that is on one of the pages that this + // large object uses. That ensures that we can use a negative result to + // populate the negative page cache. bool contains(Address object) { - return (address() <= object) && (object <= (address() + size())); + return roundToBlinkPageStart(address()) <= object && object < roundToBlinkPageEnd(address() + size()); } LargeHeapObject<Header>* next() @@ -236,7 +253,6 @@ public: void finalize(); private: - friend class Heap; friend class ThreadHeap<Header>; LargeHeapObject<Header>* m_next; @@ -439,10 +455,14 @@ public: bool isEmpty(); + // Returns true for the whole blinkPageSize page that the page is on, even + // for the header, and the unmapped guard page at the start. That ensures + // the result can be used to populate the negative page cache. bool contains(Address addr) { Address blinkPageStart = roundToBlinkPageStart(address()); - return blinkPageStart <= addr && (blinkPageStart + blinkPageSize) > addr; + ASSERT(blinkPageStart = address() - osPageSize()); // Page is at aligned address plus guard page size. + return blinkPageStart <= addr && addr < blinkPageStart + blinkPageSize; } HeapPage* next() { return m_next; } @@ -464,7 +484,7 @@ public: void sweep(); void clearObjectStartBitMap(); void finalize(Header*); - virtual bool checkAndMarkPointer(Visitor*, Address); + virtual void checkAndMarkPointer(Visitor*, Address) OVERRIDE; #if ENABLE(GC_TRACING) const GCInfo* findGCInfo(Address) OVERRIDE; #endif @@ -488,77 +508,115 @@ protected: friend class ThreadHeap<Header>; }; -// A HeapContainsCache provides a fast way of taking an arbitrary +class AddressEntry { +public: + AddressEntry() : m_address(0) { } + + explicit AddressEntry(Address address) : m_address(address) { } + + Address address() const { return m_address; } + +private: + Address m_address; +}; + +class PositiveEntry : public AddressEntry { +public: + PositiveEntry() + : AddressEntry() + , m_containingPage(0) + { + } + + PositiveEntry(Address address, BaseHeapPage* containingPage) + : AddressEntry(address) + , m_containingPage(containingPage) + { + } + + BaseHeapPage* result() const { return m_containingPage; } + + typedef BaseHeapPage* LookupResult; + +private: + BaseHeapPage* m_containingPage; +}; + +class NegativeEntry : public AddressEntry { +public: + NegativeEntry() : AddressEntry() { } + + NegativeEntry(Address address, bool) : AddressEntry(address) { } + + bool result() const { return true; } + + typedef bool LookupResult; +}; + +// A HeapExtentCache provides a fast way of taking an arbitrary // pointer-sized word, and determining whether it can be interpreted // as a pointer to an area that is managed by the garbage collected // Blink heap. There is a cache of 'pages' that have previously been -// determined to be either wholly inside or wholly outside the -// heap. The size of these pages must be smaller than the allocation -// alignment of the heap pages. We determine on-heap-ness by rounding -// down the pointer to the nearest page and looking up the page in the -// cache. If there is a miss in the cache we ask the heap to determine -// the status of the pointer by iterating over all of the heap. The -// result is then cached in the two-way associative page cache. +// determined to be wholly inside the heap. The size of these pages must be +// smaller than the allocation alignment of the heap pages. We determine +// on-heap-ness by rounding down the pointer to the nearest page and looking up +// the page in the cache. If there is a miss in the cache we can ask the heap +// to determine the status of the pointer by iterating over all of the heap. +// The result is then cached in the two-way associative page cache. // -// A HeapContainsCache is both a positive and negative -// cache. Therefore, it must be flushed both when new memory is added -// and when memory is removed from the Blink heap. -class HeapContainsCache { +// A HeapContainsCache is a positive cache. Therefore, it must be flushed when +// memory is removed from the Blink heap. The HeapDoesNotContainCache is a +// negative cache, so it must be flushed when memory is added to the heap. +template<typename Entry> +class HeapExtentCache { public: - HeapContainsCache(); + HeapExtentCache() + : m_entries(adoptArrayPtr(new Entry[HeapExtentCache::numberOfEntries])) + , m_hasEntries(false) + { + } void flush(); bool contains(Address); + bool isEmpty() { return !m_hasEntries; } // Perform a lookup in the cache. // - // If lookup returns false the argument address was not found in + // If lookup returns null/false the argument address was not found in // the cache and it is unknown if the address is in the Blink // heap. // - // If lookup returns true the argument address was found in the - // cache. In that case, the address is in the heap if the base - // heap page out parameter is different from 0 and is not in the - // heap if the base heap page out parameter is 0. - bool lookup(Address, BaseHeapPage**); - - // Add an entry to the cache. Use a 0 base heap page pointer to - // add a negative entry. - void addEntry(Address, BaseHeapPage*); - -private: - class Entry { - public: - Entry() - : m_address(0) - , m_containingPage(0) - { - } - - Entry(Address address, BaseHeapPage* containingPage) - : m_address(address) - , m_containingPage(containingPage) - { - } - - BaseHeapPage* containingPage() { return m_containingPage; } - Address address() { return m_address; } + // If lookup returns true/a page, the argument address was found in the + // cache. For the HeapContainsCache this means the address is in the heap. + // For the HeapDoesNotContainCache this means the address is not in the + // heap. + … (message too large)
The CQ bit was checked by erik.corry@gmail.com
CQ is trying da patch. Follow status at https://chromium-status.appspot.com/cq/erik.corry@gmail.com/271703002/60001
Message was sent while issue was closed.
Change committed as 173749 |